t********k
发帖数: 808 | 1 今天在一Mainframe机上进行了一个测试
增加CPU个数和增大内存大小
发现有些select语句执行速度可以大大地加快
而有些selct语句执行速度可以说根本未变
（可能还会慢上个几秒）
有些select语句只会用一颗CPU(该颗CPU使用率在100%)，
其它的空闲着
有些select语句会用所有的CPU,但使用率可能在25%
这是为什么呢？
测试环境如下：
分别用4,8,12,16颗英特尔PIII至强900MHz/2MB CPU
内存大小分别是：4GB和8GB
OS：Win2000 Data Center
RDBMS: MS SQL Server 2000 |
|
t*v
发帖数: 8 | 2 No, SAS dataset is not relational database. It has its own format. |
|
n********a
发帖数: 68 | 3 Use a *REAL* RDBMS that can grant/revoke by roles
and has row-level security and dump MS-ACCESS crap? |
|
n********a
发帖数: 68 | 4
Are your SQL statements vastly different or just
different in the variables to bind as I have
previously illustrated?
If your SQLs are very different such that you
cannot use variable binding, then that just
says you don't have enough shared pool to cache
all the very different SQLs. You want to size
up your shared pool, NOT large pool.
But again, try very very very hard to do variable
binding.
I don't know about other RDBMS, but for Oracle, it always caches
SQL. |
|
n********a
发帖数: 68 | 5
Oracle used to dominate in terms of marketshare
and was M$ in database area. However, it now
faces fierce competition from M$'s SQLServer
and IBM's DB2 as well as Open source choices.
Technically, Oracle is a better RDBMS than
other commercial or open source products, IMHO.
But $ speaks much louder. |
|
n********a
发帖数: 68 | 6 Excel, are you serious?
Do you know that Excel has a limit of 65536 of rows?
Use a real RDBMS or at least Access.
平
警
低 |
|
w*r
发帖数: 2421 | 7 data warehouse要什么教程?????data warehouse实际上不是一个technology term, 是
一个business term. 本质上都是RDBMS, 关键是什么样的application数据库就是那么
点东西,随便找本书看就完了 |
|
WB
发帖数: 170 | 8 比如oracle7/8
sybase5/9/11
informix 7/9
这些在现在的新硬件服务器 上oltp/rdbms的性能（查询，插入删除 增加。。。
，存储过程）
跟他们新版本比较不会差个50%-200%吧？ |
|
w*r
发帖数: 2421 | 9 mysql...well, free software usually need lot tuning effort. I believe mysql
need some tuning to make the db engine choose hash join on index. Other
commercial db software should use merge or hash join for the index
automatically in your scenario. get away from mysql if you do not want to
waste your time to fine tune the RDBMS behavior.
(A |
|
w*r
发帖数: 2421 | 10 我当然知道在oracle里面这个insert会被变成null, 这正是问题所在,TNND 这太烦了,
老爷偶要null的时候当然会把null给送去表里,你丫给我自动换什么呀，这个行为几乎
和所有其他的RDBMS都不一样！
造成的结果是如果有一个java function embeded在数据库里面,送去一个空字符的时候
就会跳出null exception, 你说丫的是不是很烦? |
|
t**********t
发帖数: 3 | 11 I don't think any RDBMS have the built-in capability to read a random .sql
file and "import" it. You most likely will have to write code to do that. |
|
q**1
发帖数: 193 | 12 我看你应该吧你的问题细分下：
* 关于结合 RDBMS 和 XML and its friends的相关技术，
how and what you expect?（技术问题）
* SleepyCat 和 Oracle 的关系（历史问题）。
* 两种数据库之间的不同，彼此的优缺点（技术问题）。
google吧，呵呵。。 |
|
|
w*r
发帖数: 2421 | 14 build a tree diagram of your query you will find that you have several
tables scanned multiple times
i re-write your query by eliminating the IN() construct. Usually the IN
construct is not well optimized in most of RDBMS system. In addition to this
, I believe you can use OLAP MAX() PARTITION BY() function to eliminate the
cs_of_date subquery.
SELECT
DECODE (LENGTH (e.e_last_name),0, u.user_id, e.e_last_name || ', ' || e.
e_fi
rst_name) salesperson_name,
t.ca_care_id AS account_numb |
|
j*****n
发帖数: 1781 | 15
> 1GB memory is recommended, otherwise most PCs are ok.
xp and Srv2003
Get some DB fundamental books first to have the basic knowledge of RDBMS.
Then SQL Server Online books (help file) is your best friend. It provides
every single details for you. |
|
|
B*****g
发帖数: 34098 | 17 no diff?
One is for # of OS, one is for Revenue |
|
w*r
发帖数: 2421 | 18 not exists在oracle/db2/teradata 上面优化得都很烂，netezza还成，
SQL server这种没有optimizer的RDBMS不谈也罢 |
|
w*********e
发帖数: 5286 | 19 the kernal Oracle RDBMS are all in C, from parser to execution layer. |
|
v***e
发帖数: 2108 | 20 I really doubt it would happen. MySQL does not compete directly with Oracle
RDBMS in some sense, it competes more with MS SQL.
After mysql's reunion with InnoDB in Oracle,
it could be a better weapon against mssql. |
|
v*****r
发帖数: 1119 | 21 配置说明：
2 nodes Oracle RAC using ASM with OCR/vote disks on raw devices
Clusterware: 10.2.0.4
ASM: 10.2.0.4
RDBMS: 10.2.0.4
Operating System: RedHat5.3
Storage: SAN |
|
v*****r
发帖数: 1119 | 22 写的不错，补充一下:
Isolation level 是ANSI SQL 的标准，不是SQL Server 的概念。每个 rdbms 实现
Isolation level 的方式都不一样，这也是 Oracle 比 SQL Server 设计上更优异的
一个地方。举个简单的例子，Oracle 的 Undo image的概念保证了 non-blocking read
没有 dirty read 的忧虑，所以根本不需要象 SQL Sever 那样为了提高 performance
牺牲 integrity。换句话说，同样的 transaction isolation level 下 (by default
都是 Read Committed), Oracle 不需要 SQL Server 那么多 Locking，所以 Oracle
在 performance 上比 SQL Server 有先天的优势。
前几天被老板逼着去 Redmond 上了MS SQLServer 两周课，认真做了 one on one 的比
较，不可否认，SQL Server 也有些相对的优点，不过在很多地方和 Oracl |
|
v*****r
发帖数: 1119 | 23 They can be used together with group/having, but not inside group/having.
The reason is analytic functions always the last step to be computed (except
ORDER BY clause), so only legal to be put in SELECT/ORDER BY.
Almost all analytic functions are designed by Oracle first, then got
accepted into SQL89/92 standards, so they should behave the same for all
rdbms including SQLSerer.
in |
|
d*******o
发帖数: 493 | 24 商务智能，数据挖掘，RDBMS不适合，SQL的优点是interactive mode，对这个没什么帮
助。Web-scale的数据，不管MySQL还是Oracle sort一些下都歇菜了，别提做training
和scoring了。
传统的DBA可以转到Hadoop上面的Hive上去。新生代的DBA直接用java/python/ruby 做
map和reduce的streaming。 |
|
y****9
发帖数: 144 | 25 >> 现在DBA看report是run自己的script还是去EM上看？
I think it really depends on different situations. Being able to run scripts
only from command-line is a must-have skill, and it works for all situation
and across verisons and different shops.
Not every shop has GUI tool. we don't have EM at this moment
>> 感觉oracle也向傻瓜型发展。
In terms of managability of RDBMS software or their marketing goal at least,
it seems to be true.
But I don't think an Oracle DBA can be challenged to any less extent by less
number of is... 阅读全帖 |
|
c*****t
发帖数: 1879 | 26 RDBMS is for structured (not hierarchical, but cleaned) data.
Schema allows efficient query planning.
For unstructured data, use NoSQL databases. |
|
|
v***e
发帖数: 2108 | 28 我觉得这种说法在以前是可以理解，但是现在未必是这样
interview倒也罢了，但是如果是实际工作，XML这种东西一
定比pl/sql快么？ 好像很难说吧。
写一个SQL statment并不能简化solution的内在逻辑，只是把所有
东西压缩在一个stmt里面了。
而且即使是写一个巨大，复杂，难以维护的single SQL statement，
也未必一定比一个简单，逻辑清晰的compiled pl/sql function
更efficient，其中一个原因是一旦你SQL stmt复杂之后，plan
generation 和plan execution都很难做到opitimized。而且复杂的sql
像cursor sharing，result cache这种feature就更难运用。
从RDBMS内核的角度看，我个人不太赞成把任何东西都写在一个SQL里面。 |
|
y****w
发帖数: 3747 | 29 google:
Can one improve the performance of SQL*Loader?
A very simple but easily overlooked hint is not to have any indexes and/
or constraints (primary key) on your load tables during the load process.
This will significantly slow down load times even with ROWS= set to a high
value.
Add the following option in the command line: DIRECT=TRUE. This will
effectively bypass most of the RDBMS processing. However, there are cases
when you can't use direct load. For details, refer to the FAQ abo... 阅读全帖 |
|
M*****r
发帖数: 1536 | 30 我刚刚接触一些，发现和“传统”RDBMS非常不同（它们内部不同实现之间大概也有差
别）
column-based
column-compression
no index
MPP
many are postgres based (Aster/Vertica/Greenplum/Netezza |
|
a*********8
发帖数: 9 | 31 mpp analytic db, not "traditional rdbms", but not nosql either |
|
g***l
发帖数: 18555 | 32 Increase productivity and save businesses up to $58,800 (USD) per year per
DBA by using Oracle Database 11g over Microsoft SQL Server 2008
这个扯吧，ORACLE DBA工资比SQL SERVER高个3万5万的，而且OCP们根本不搞开发，连
个FEED什么的都不管，就这还省呢。 |
|
B*****g
发帖数: 34098 | 33 oracle DBA比sql server DBA挣的少 |
|
|
g***l
发帖数: 18555 | 35 打死我都不信。ORACLE DBA至少比SQL SERVER高2万，我当初就因为这想转ORACLE，人嫌我没经验。ORACLE起点高，挣得高，SQL SERVER好多公司就穷糊弄，4万就招个DBA。 |
|
|
B*****g
发帖数: 34098 | 37 去看我贴的数据，DBA比developer挣的还少 |
|
g***l
发帖数: 18555 | 38 我不信那些数据，我这SQL DEVELOPER加州的，5万有干的，DBA7，8万都招不到人 |
|
|
g***l
发帖数: 18555 | 40 就这价，爱干不干。也有给的高的，挤破门。这边中高级的SQL SERVER DBA现在比较缺
，水平稍微好一点的都比较少见，所以我觉得那些入门的应该把起点瞄得高一点，
PAPER DBA也行的。 |
|
|
w*r
发帖数: 2421 | 42 the keypoint is make use . the research and analytic work usually do not
know exactly what he/she needs when exploring the data. the specific need
and analytics methodology did not materialize until very late in the game of
data exploring and standardizing. RDBMS still monopolize the field of data
management. Hadoop framework will provide a supplement platform to produce
the analytics computing power to replace traditional cube/statistic/modeling
engine between DW and data consumers.
the
cloud... 阅读全帖 |
|
l*******G
发帖数: 1191 | 43 thanks for sharing. Like simple arrays are much more efficient to manage
than dynamic graphs in memory, RDBMS (tables) are much easier and more
efficient in managing/indexing data. Managing of unstructured data should be
avoided as much as possible. From this point of view, I don't see big
future with big unstructured data. Only when cost in managing unstructured
data is lowered to be close to cost of managing tables, can I see its
benefit in application.
Secondly, when pepople say unstru... 阅读全帖 |
|
w*r
发帖数: 2421 | 44 【 以下文字转载自 Texas 讨论区 】
发信人: wyr (遗忘小资--其实我是挖坑的), 信区: Texas
标 题: Enterprise data warehouse team is looking for developers and senior developers
发信站: BBS 未名空间站 (Fri Jan 18 12:05:09 2013, 美东)
Location: Irving Texas.
Industry: healthcare.
Requirements:
Understands the RDBMS architect and practical knowledge of using SQL to
extract/transform/analyze data.
Good analytical skills .
SQL development ANSI
teradata experience is abig plus.
Unix shell scripting
Perl is a plus.
good communication skills
ETL ... 阅读全帖 |
|
w*r
发帖数: 2421 | 45 please check google GFS. google.baidu.taobao.fb definitely are not using
traditional RDBMS to store./search/process data. google has its own google
file sys, taobao has its own taobao FS.. i believe baidu and others are
pursuing the same route |
|
F****n
发帖数: 3271 | 46 State of the Arts 是啥？
比如可以同时N个读写的B-tree之类的 |
|
c*********e
发帖数: 16335 | 47 ibm db2咋样了，ibm内部都在用db2, lotusnotes.
oracle毕竟发展得早，专门搞database的，比microsoft搞database早多了。larry
ellison又那么精明，现在没结婚了，有更多时间放工作上了。
combined |
|
w*r
发帖数: 2421 | 48 hate oracle hate oracl hate oracle
period....
combined |
|
|
c*********e
发帖数: 16335 | 50 lotus notes隐私比gmail好，使用功能比outlook多。好多大公司都用它。 |
|
